Update README.md

Author: hamelsmu

README.md

@@ -4,8 +4,26 @@ GitHub contains a large corpus data that is amenable for NLP, in the form of Iss
 
 ## 1. Insert custom field indicators
 
-This is so markdown information is not lost.  For example, a list block is enclosed with `xxxlistB` and `xxxlistE` and a code block is enclosed with `xxxcdb` and `xxxcde`.  
+This is so markdown information is not lost.  For example, a list block is enclosed with `xxxlistB` and `xxxlistE` and a code block is enclosed with `xxxcdb` and `xxxcde`.  Other noteable examples:
+
+- @mentions:  xxxatmention  (the handle is removed and replaced by just this indicator)
+- quote blocks: xxxqb/xxxqe
+- strikethrough: xxxdelb/xxxdele
+- horizontal rule: xxxhr
+- {large, medium, small} headers: annotated with xxxh{l,m,s}.  H1=large, H2-3=medium, H4-6=small.
+-
 
 ## 2. Discard superflous information
 
-Documentation TBD
+GitHub issues often contain a large stack trace, or a large table with data.  This library comes equipped with sensible defaults to surface the most relevant information and discard what would otherwhise be lots of characters for a machine learning algorithm to handle:
+
+- Code Blocks: only first two and last two rows are kept 
+- Tables:  only table headers are kept
+- Urls:  only the host is kept.  For example www.google.com/search is reformatted to www.google.com
+- Images: the image is discarded but the file extension and metadata about the image (available to screenreader) is extracted.
+- IP Addresses, extremely long numbers are marked as xxunk
+
+
+# Examples
+
+See [/notebooks/Demo.ipynb](/notebooks/Demo.ipynb) for an example of the transformations this parser does on a markdown file.